Online Learning to Rank for Recommender Systems
by Daan Odijk (Blendle)
Every morning, at Blendle, we have a huge cold-start problem when over 8.000 new articles from the latest editions of newspapers arrive in our system. At that moment, these articles are read by virtually no-one and we are tasked with sending out personalised newsletters to over 1 million users. We can thus not rely on collaborative filtering type of recommendations, nor can we use the popularity of the articles as clues for what our user might want to read. We overcome our cold-start problem by a mix of curation by our editorial team and an automated analysis of the content of these articles. We extract named entities, semantic links, authors, the language and plenty of stylometrics. For each of our users, we build a very fine grained profile based on the attributes of the articles that they read. The combination of enriched articles and user profiles is fed into our machine learning pipeline. We are currently experimenting with an online learning to rank setup, where each of our users is exposed to a slightly perturbed version of our ranking model. We observe the interactions of our users to infer in which direction we should be updating the model.
Our editorial team gets up at around 5am every morning to read what was published over night. They are done reading and recommending their selection of articles around 8am, which is also the time we would ideally send out the newsletter so that our users, on their commute to work, can read our newsletter. These timing restrictions pose yet another challenge: our content analysis and machine learning pipeline needs to be really fast. We solve this by using a streaming infrastructure build on Kafka. In this infrastructure, an article is analysed and scored for relevance towards each of our users as soons as it arrives. This has the advantage that at 8am, when our editorial team is done reading, personalisation is much more lightweight. We use the precomputed relevance scores and balance them with diversity to arrive at a unique ranking for each of our users. In this talk, I will detail how we enrich articles in a streaming fashion and how we use online learning methods to learn a ranking model. I will also talk about how we deal with the time constraints of the problem we are trying to solve.
About the Speaker
Daan Odijk is lead data scientist at Blendle, a New York Times backed startup that builds a platform where users can explore and support the world’s best journalism and only pay for what they read. Daan heads a team of eight data scientists and engineers who work on personalised recommendations. Daan has a PhD in information retrieval and has worked on leveraging context when searching for news.
Slides
https://www.slideshare.net/…/blendle-recsys17-online-learning-to-rank-for-recommender-systems
Bandit Algorithms for e-Commerce Recommender Systems
by Mikael Hammar (Apptus)
We study bandit algorithms for e-commerce recommender systems. The question we pose is whether it is necessary to consider reinforcement learning effects in recommender systems. A key reason to introduce a recommender system for a product page on an e-commerce site is to increase the order value by improving the chance of making an upsale. If the recommender system merely predicts the next purchase, there might be no positive effect at all on the order value, since the recommender system predicts sales that would have happened independent of the recommender system. What we really are looking for are the false negatives, i.e., purchases that happen as a consequence of the recommender system. These purchases entail the entire uplift and should be present as reinforcement learning effects. This effect cannot be displayed in a simulation of the site, since there are no reinforcement learning effects present in a simulation. The attribution model must capture the uplift to guarantee an increased order value. However, such an attribution model is not practical, due to data sparsity. Given this starting point, we study some standard attribution models for e-commerce recommender systems, and describe how these fare when applied in a reinforcement learning algorithm, both in a simulation and on live sites.
About the Speaker
Head of Research at Apptus Technologies AB, a Swedish tech-company that provides services that facilitate search, navigation and recommendations on high-end e-commerce sites. Mikael Hammar took his PhD in computer science at Lund University, and is specialized in approximation algorithms for NP-hard problems.
Slides
https://www.slideshare.net/…/bandit-algorithms-for-ecommerce-recommender-systems
Transfer Learning for Personalized Content and Ad Recommendation
by Zhixian Yan (Cheetah Mobile) and Lai Wei (Cheetah Mobile)
Cold start is always a key challenge for building real-life recommendation systems. Thanks to the ever-growing multi-modal data in the mobile Internet age and the latest deep learning techniques, transfer-learning based cross-domain recommendation starts to play a crucial role in tackling the cold start problem and to provide “warm-start” recommendation for new users.
In this talk, we will introduce the experiences and lessons that we have learnt from building personalized recommendation systems for both advertisement and content scenarios at Cheetah Mobile, serving 600+ millions monthly active mobile users. In particular, we leveraged the app install & usage and many other mobile data, built a Unified User Profile (UUP) by using transfer learning and deep learning, and developed cross-domain personalized Ad and news recommendation. Our approaches enable us to solve the cold start problem with close to full coverage of our user base while yielding significant CTR increase and better user experience.
About the Speakers
Zhixian Yan is a Senior Staff Research Engineer at Cheetah Mobile, leading the data and algorithm perspectives on several AI projects. Previously, Zhixian was the first and principal data scientist at GoDaddy Inc., and a staff research engineer at Samsung Research America. He received his Ph.D. from EPFL (Swiss Federal Institute of Technology – Lausanne) and M.Sc. from CAS (Chinese Academy of Science). Zhixian has published 30+ papers with 1400+ citations. His current research and engineering interests lie in building real-life large-scale personalized recommendation systems, deep learning, and all kinds of AI applications.
Lai Wei is a Software Engineer at Cheetah Mobile, working on big data pipelines and machine learning/deep learning algorithms. Lai received his M.Sc. in Electrical and Computer Engineering from Carnegie Mellon University (CMU) and B.Eng. in Electrical & Electronic Engineering from Nanyang Technological University. He is interested in deep learning, user profiling, and recommendation systems.
